Abstract: To accurately predict wind power, this study proposes a short-term wind power prediction model based on complementary ensemble empirical modal decomposition (CEEMD), improved snake optimization algorithm (ISO), and kernel extreme learning machine (KELM). Firstly, the non-smooth wind power data are decomposed into multiple relatively smoother components using CEEMD to reduce the instability of the original data. Then, an improved snake optimization algorithm is introduced to optimize the parameters of KELM, and the CEEMD-ISO-KELM prediction model is constructed for each smooth component and residual. Finally, the prediction results of each component and residual are reconstructed to obtain the final wind power prediction results. The simulation results demonstrate that, in comparison with existing prediction models, the proposed model in this study exhibits exceptional capability in accurately predicting short-term wind power trends.